2 * (C) 1997 Linus Torvalds
3 * (C) 1999 Andrea Arcangeli <andrea@suse.de> (dynamic inode allocation)
7 #include <linux/dcache.h>
8 #include <linux/init.h>
9 #include <linux/slab.h>
10 #include <linux/writeback.h>
11 #include <linux/module.h>
12 #include <linux/backing-dev.h>
13 #include <linux/wait.h>
14 #include <linux/rwsem.h>
15 #include <linux/hash.h>
16 #include <linux/swap.h>
17 #include <linux/security.h>
18 #include <linux/pagemap.h>
19 #include <linux/cdev.h>
20 #include <linux/bootmem.h>
21 #include <linux/fsnotify.h>
22 #include <linux/mount.h>
23 #include <linux/async.h>
24 #include <linux/posix_acl.h>
25 #include <linux/prefetch.h>
26 #include <linux/ima.h>
27 #include <linux/cred.h>
28 #include <linux/buffer_head.h> /* for inode_has_buffers */
32 * Inode locking rules:
34 * inode->i_lock protects:
35 * inode->i_state, inode->i_hash, __iget()
36 * inode->i_sb->s_inode_lru_lock protects:
37 * inode->i_sb->s_inode_lru, inode->i_lru
38 * inode_sb_list_lock protects:
39 * sb->s_inodes, inode->i_sb_list
40 * bdi->wb.list_lock protects:
41 * bdi->wb.b_{dirty,io,more_io}, inode->i_wb_list
42 * inode_hash_lock protects:
43 * inode_hashtable, inode->i_hash
49 * inode->i_sb->s_inode_lru_lock
62 static unsigned int i_hash_mask __read_mostly
;
63 static unsigned int i_hash_shift __read_mostly
;
64 static struct hlist_head
*inode_hashtable __read_mostly
;
65 static __cacheline_aligned_in_smp
DEFINE_SPINLOCK(inode_hash_lock
);
67 __cacheline_aligned_in_smp
DEFINE_SPINLOCK(inode_sb_list_lock
);
70 * Empty aops. Can be used for the cases where the user does not
71 * define any of the address_space operations.
73 const struct address_space_operations empty_aops
= {
75 EXPORT_SYMBOL(empty_aops
);
78 * Statistics gathering..
80 struct inodes_stat_t inodes_stat
;
82 static DEFINE_PER_CPU(unsigned int, nr_inodes
);
83 static DEFINE_PER_CPU(unsigned int, nr_unused
);
85 static struct kmem_cache
*inode_cachep __read_mostly
;
87 static int get_nr_inodes(void)
91 for_each_possible_cpu(i
)
92 sum
+= per_cpu(nr_inodes
, i
);
93 return sum
< 0 ? 0 : sum
;
96 static inline int get_nr_inodes_unused(void)
100 for_each_possible_cpu(i
)
101 sum
+= per_cpu(nr_unused
, i
);
102 return sum
< 0 ? 0 : sum
;
105 int get_nr_dirty_inodes(void)
107 /* not actually dirty inodes, but a wild approximation */
108 int nr_dirty
= get_nr_inodes() - get_nr_inodes_unused();
109 return nr_dirty
> 0 ? nr_dirty
: 0;
113 * Handle nr_inode sysctl
116 int proc_nr_inodes(ctl_table
*table
, int write
,
117 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
119 inodes_stat
.nr_inodes
= get_nr_inodes();
120 inodes_stat
.nr_unused
= get_nr_inodes_unused();
121 return proc_dointvec(table
, write
, buffer
, lenp
, ppos
);
126 * inode_init_always - perform inode structure intialisation
127 * @sb: superblock inode belongs to
128 * @inode: inode to initialise
130 * These are initializations that need to be done on every inode
131 * allocation as the fields are not initialised by slab allocation.
133 int inode_init_always(struct super_block
*sb
, struct inode
*inode
)
135 static const struct inode_operations empty_iops
;
136 static const struct file_operations empty_fops
;
137 struct address_space
*const mapping
= &inode
->i_data
;
140 inode
->i_blkbits
= sb
->s_blocksize_bits
;
142 atomic_set(&inode
->i_count
, 1);
143 inode
->i_op
= &empty_iops
;
144 inode
->i_fop
= &empty_fops
;
148 atomic_set(&inode
->i_writecount
, 0);
152 inode
->i_generation
= 0;
154 memset(&inode
->i_dquot
, 0, sizeof(inode
->i_dquot
));
156 inode
->i_pipe
= NULL
;
157 inode
->i_bdev
= NULL
;
158 inode
->i_cdev
= NULL
;
160 inode
->dirtied_when
= 0;
162 if (security_inode_alloc(inode
))
164 spin_lock_init(&inode
->i_lock
);
165 lockdep_set_class(&inode
->i_lock
, &sb
->s_type
->i_lock_key
);
167 mutex_init(&inode
->i_mutex
);
168 lockdep_set_class(&inode
->i_mutex
, &sb
->s_type
->i_mutex_key
);
170 atomic_set(&inode
->i_dio_count
, 0);
172 mapping
->a_ops
= &empty_aops
;
173 mapping
->host
= inode
;
175 mapping_set_gfp_mask(mapping
, GFP_HIGHUSER_MOVABLE
);
176 mapping
->assoc_mapping
= NULL
;
177 mapping
->backing_dev_info
= &default_backing_dev_info
;
178 mapping
->writeback_index
= 0;
181 * If the block_device provides a backing_dev_info for client
182 * inodes then use that. Otherwise the inode share the bdev's
186 struct backing_dev_info
*bdi
;
188 bdi
= sb
->s_bdev
->bd_inode
->i_mapping
->backing_dev_info
;
189 mapping
->backing_dev_info
= bdi
;
191 inode
->i_private
= NULL
;
192 inode
->i_mapping
= mapping
;
193 #ifdef CONFIG_FS_POSIX_ACL
194 inode
->i_acl
= inode
->i_default_acl
= ACL_NOT_CACHED
;
197 #ifdef CONFIG_FSNOTIFY
198 inode
->i_fsnotify_mask
= 0;
201 this_cpu_inc(nr_inodes
);
207 EXPORT_SYMBOL(inode_init_always
);
209 static struct inode
*alloc_inode(struct super_block
*sb
)
213 if (sb
->s_op
->alloc_inode
)
214 inode
= sb
->s_op
->alloc_inode(sb
);
216 inode
= kmem_cache_alloc(inode_cachep
, GFP_KERNEL
);
221 if (unlikely(inode_init_always(sb
, inode
))) {
222 if (inode
->i_sb
->s_op
->destroy_inode
)
223 inode
->i_sb
->s_op
->destroy_inode(inode
);
225 kmem_cache_free(inode_cachep
, inode
);
232 void free_inode_nonrcu(struct inode
*inode
)
234 kmem_cache_free(inode_cachep
, inode
);
236 EXPORT_SYMBOL(free_inode_nonrcu
);
238 void __destroy_inode(struct inode
*inode
)
240 BUG_ON(inode_has_buffers(inode
));
241 security_inode_free(inode
);
242 fsnotify_inode_delete(inode
);
243 #ifdef CONFIG_FS_POSIX_ACL
244 if (inode
->i_acl
&& inode
->i_acl
!= ACL_NOT_CACHED
)
245 posix_acl_release(inode
->i_acl
);
246 if (inode
->i_default_acl
&& inode
->i_default_acl
!= ACL_NOT_CACHED
)
247 posix_acl_release(inode
->i_default_acl
);
249 this_cpu_dec(nr_inodes
);
251 EXPORT_SYMBOL(__destroy_inode
);
253 static void i_callback(struct rcu_head
*head
)
255 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
256 INIT_LIST_HEAD(&inode
->i_dentry
);
257 kmem_cache_free(inode_cachep
, inode
);
260 static void destroy_inode(struct inode
*inode
)
262 BUG_ON(!list_empty(&inode
->i_lru
));
263 __destroy_inode(inode
);
264 if (inode
->i_sb
->s_op
->destroy_inode
)
265 inode
->i_sb
->s_op
->destroy_inode(inode
);
267 call_rcu(&inode
->i_rcu
, i_callback
);
270 void address_space_init_once(struct address_space
*mapping
)
272 memset(mapping
, 0, sizeof(*mapping
));
273 INIT_RADIX_TREE(&mapping
->page_tree
, GFP_ATOMIC
);
274 spin_lock_init(&mapping
->tree_lock
);
275 mutex_init(&mapping
->i_mmap_mutex
);
276 INIT_LIST_HEAD(&mapping
->private_list
);
277 spin_lock_init(&mapping
->private_lock
);
278 INIT_RAW_PRIO_TREE_ROOT(&mapping
->i_mmap
);
279 INIT_LIST_HEAD(&mapping
->i_mmap_nonlinear
);
281 EXPORT_SYMBOL(address_space_init_once
);
284 * These are initializations that only need to be done
285 * once, because the fields are idempotent across use
286 * of the inode, so let the slab aware of that.
288 void inode_init_once(struct inode
*inode
)
290 memset(inode
, 0, sizeof(*inode
));
291 INIT_HLIST_NODE(&inode
->i_hash
);
292 INIT_LIST_HEAD(&inode
->i_dentry
);
293 INIT_LIST_HEAD(&inode
->i_devices
);
294 INIT_LIST_HEAD(&inode
->i_wb_list
);
295 INIT_LIST_HEAD(&inode
->i_lru
);
296 address_space_init_once(&inode
->i_data
);
297 i_size_ordered_init(inode
);
298 #ifdef CONFIG_FSNOTIFY
299 INIT_HLIST_HEAD(&inode
->i_fsnotify_marks
);
302 EXPORT_SYMBOL(inode_init_once
);
304 static void init_once(void *foo
)
306 struct inode
*inode
= (struct inode
*) foo
;
308 inode_init_once(inode
);
312 * inode->i_lock must be held
314 void __iget(struct inode
*inode
)
316 atomic_inc(&inode
->i_count
);
320 * get additional reference to inode; caller must already hold one.
322 void ihold(struct inode
*inode
)
324 WARN_ON(atomic_inc_return(&inode
->i_count
) < 2);
326 EXPORT_SYMBOL(ihold
);
328 static void inode_lru_list_add(struct inode
*inode
)
330 spin_lock(&inode
->i_sb
->s_inode_lru_lock
);
331 if (list_empty(&inode
->i_lru
)) {
332 list_add(&inode
->i_lru
, &inode
->i_sb
->s_inode_lru
);
333 inode
->i_sb
->s_nr_inodes_unused
++;
334 this_cpu_inc(nr_unused
);
336 spin_unlock(&inode
->i_sb
->s_inode_lru_lock
);
339 static void inode_lru_list_del(struct inode
*inode
)
341 spin_lock(&inode
->i_sb
->s_inode_lru_lock
);
342 if (!list_empty(&inode
->i_lru
)) {
343 list_del_init(&inode
->i_lru
);
344 inode
->i_sb
->s_nr_inodes_unused
--;
345 this_cpu_dec(nr_unused
);
347 spin_unlock(&inode
->i_sb
->s_inode_lru_lock
);
351 * inode_sb_list_add - add inode to the superblock list of inodes
352 * @inode: inode to add
354 void inode_sb_list_add(struct inode
*inode
)
356 spin_lock(&inode_sb_list_lock
);
357 list_add(&inode
->i_sb_list
, &inode
->i_sb
->s_inodes
);
358 spin_unlock(&inode_sb_list_lock
);
360 EXPORT_SYMBOL_GPL(inode_sb_list_add
);
362 static inline void inode_sb_list_del(struct inode
*inode
)
364 spin_lock(&inode_sb_list_lock
);
365 list_del_init(&inode
->i_sb_list
);
366 spin_unlock(&inode_sb_list_lock
);
369 static unsigned long hash(struct super_block
*sb
, unsigned long hashval
)
373 tmp
= (hashval
* (unsigned long)sb
) ^ (GOLDEN_RATIO_PRIME
+ hashval
) /
375 tmp
= tmp
^ ((tmp
^ GOLDEN_RATIO_PRIME
) >> i_hash_shift
);
376 return tmp
& i_hash_mask
;
380 * __insert_inode_hash - hash an inode
381 * @inode: unhashed inode
382 * @hashval: unsigned long value used to locate this object in the
385 * Add an inode to the inode hash for this superblock.
387 void __insert_inode_hash(struct inode
*inode
, unsigned long hashval
)
389 struct hlist_head
*b
= inode_hashtable
+ hash(inode
->i_sb
, hashval
);
391 spin_lock(&inode_hash_lock
);
392 spin_lock(&inode
->i_lock
);
393 hlist_add_head(&inode
->i_hash
, b
);
394 spin_unlock(&inode
->i_lock
);
395 spin_unlock(&inode_hash_lock
);
397 EXPORT_SYMBOL(__insert_inode_hash
);
400 * remove_inode_hash - remove an inode from the hash
401 * @inode: inode to unhash
403 * Remove an inode from the superblock.
405 void remove_inode_hash(struct inode
*inode
)
407 spin_lock(&inode_hash_lock
);
408 spin_lock(&inode
->i_lock
);
409 hlist_del_init(&inode
->i_hash
);
410 spin_unlock(&inode
->i_lock
);
411 spin_unlock(&inode_hash_lock
);
413 EXPORT_SYMBOL(remove_inode_hash
);
415 void end_writeback(struct inode
*inode
)
419 * We have to cycle tree_lock here because reclaim can be still in the
420 * process of removing the last page (in __delete_from_page_cache())
421 * and we must not free mapping under it.
423 spin_lock_irq(&inode
->i_data
.tree_lock
);
424 BUG_ON(inode
->i_data
.nrpages
);
425 spin_unlock_irq(&inode
->i_data
.tree_lock
);
426 BUG_ON(!list_empty(&inode
->i_data
.private_list
));
427 BUG_ON(!(inode
->i_state
& I_FREEING
));
428 BUG_ON(inode
->i_state
& I_CLEAR
);
429 inode_sync_wait(inode
);
430 /* don't need i_lock here, no concurrent mods to i_state */
431 inode
->i_state
= I_FREEING
| I_CLEAR
;
433 EXPORT_SYMBOL(end_writeback
);
436 * Free the inode passed in, removing it from the lists it is still connected
437 * to. We remove any pages still attached to the inode and wait for any IO that
438 * is still in progress before finally destroying the inode.
440 * An inode must already be marked I_FREEING so that we avoid the inode being
441 * moved back onto lists if we race with other code that manipulates the lists
442 * (e.g. writeback_single_inode). The caller is responsible for setting this.
444 * An inode must already be removed from the LRU list before being evicted from
445 * the cache. This should occur atomically with setting the I_FREEING state
446 * flag, so no inodes here should ever be on the LRU when being evicted.
448 static void evict(struct inode
*inode
)
450 const struct super_operations
*op
= inode
->i_sb
->s_op
;
452 BUG_ON(!(inode
->i_state
& I_FREEING
));
453 BUG_ON(!list_empty(&inode
->i_lru
));
455 inode_wb_list_del(inode
);
456 inode_sb_list_del(inode
);
458 if (op
->evict_inode
) {
459 op
->evict_inode(inode
);
461 if (inode
->i_data
.nrpages
)
462 truncate_inode_pages(&inode
->i_data
, 0);
463 end_writeback(inode
);
465 if (S_ISBLK(inode
->i_mode
) && inode
->i_bdev
)
467 if (S_ISCHR(inode
->i_mode
) && inode
->i_cdev
)
470 remove_inode_hash(inode
);
472 spin_lock(&inode
->i_lock
);
473 wake_up_bit(&inode
->i_state
, __I_NEW
);
474 BUG_ON(inode
->i_state
!= (I_FREEING
| I_CLEAR
));
475 spin_unlock(&inode
->i_lock
);
477 destroy_inode(inode
);
481 * dispose_list - dispose of the contents of a local list
482 * @head: the head of the list to free
484 * Dispose-list gets a local list with local inodes in it, so it doesn't
485 * need to worry about list corruption and SMP locks.
487 static void dispose_list(struct list_head
*head
)
489 while (!list_empty(head
)) {
492 inode
= list_first_entry(head
, struct inode
, i_lru
);
493 list_del_init(&inode
->i_lru
);
500 * evict_inodes - evict all evictable inodes for a superblock
501 * @sb: superblock to operate on
503 * Make sure that no inodes with zero refcount are retained. This is
504 * called by superblock shutdown after having MS_ACTIVE flag removed,
505 * so any inode reaching zero refcount during or after that call will
506 * be immediately evicted.
508 void evict_inodes(struct super_block
*sb
)
510 struct inode
*inode
, *next
;
513 spin_lock(&inode_sb_list_lock
);
514 list_for_each_entry_safe(inode
, next
, &sb
->s_inodes
, i_sb_list
) {
515 if (atomic_read(&inode
->i_count
))
518 spin_lock(&inode
->i_lock
);
519 if (inode
->i_state
& (I_NEW
| I_FREEING
| I_WILL_FREE
)) {
520 spin_unlock(&inode
->i_lock
);
524 inode
->i_state
|= I_FREEING
;
525 inode_lru_list_del(inode
);
526 spin_unlock(&inode
->i_lock
);
527 list_add(&inode
->i_lru
, &dispose
);
529 spin_unlock(&inode_sb_list_lock
);
531 dispose_list(&dispose
);
535 * invalidate_inodes - attempt to free all inodes on a superblock
536 * @sb: superblock to operate on
537 * @kill_dirty: flag to guide handling of dirty inodes
539 * Attempts to free all inodes for a given superblock. If there were any
540 * busy inodes return a non-zero value, else zero.
541 * If @kill_dirty is set, discard dirty inodes too, otherwise treat
544 int invalidate_inodes(struct super_block
*sb
, bool kill_dirty
)
547 struct inode
*inode
, *next
;
550 spin_lock(&inode_sb_list_lock
);
551 list_for_each_entry_safe(inode
, next
, &sb
->s_inodes
, i_sb_list
) {
552 spin_lock(&inode
->i_lock
);
553 if (inode
->i_state
& (I_NEW
| I_FREEING
| I_WILL_FREE
)) {
554 spin_unlock(&inode
->i_lock
);
557 if (inode
->i_state
& I_DIRTY
&& !kill_dirty
) {
558 spin_unlock(&inode
->i_lock
);
562 if (atomic_read(&inode
->i_count
)) {
563 spin_unlock(&inode
->i_lock
);
568 inode
->i_state
|= I_FREEING
;
569 inode_lru_list_del(inode
);
570 spin_unlock(&inode
->i_lock
);
571 list_add(&inode
->i_lru
, &dispose
);
573 spin_unlock(&inode_sb_list_lock
);
575 dispose_list(&dispose
);
580 static int can_unuse(struct inode
*inode
)
582 if (inode
->i_state
& ~I_REFERENCED
)
584 if (inode_has_buffers(inode
))
586 if (atomic_read(&inode
->i_count
))
588 if (inode
->i_data
.nrpages
)
594 * Walk the superblock inode LRU for freeable inodes and attempt to free them.
595 * This is called from the superblock shrinker function with a number of inodes
596 * to trim from the LRU. Inodes to be freed are moved to a temporary list and
597 * then are freed outside inode_lock by dispose_list().
599 * Any inodes which are pinned purely because of attached pagecache have their
600 * pagecache removed. If the inode has metadata buffers attached to
601 * mapping->private_list then try to remove them.
603 * If the inode has the I_REFERENCED flag set, then it means that it has been
604 * used recently - the flag is set in iput_final(). When we encounter such an
605 * inode, clear the flag and move it to the back of the LRU so it gets another
606 * pass through the LRU before it gets reclaimed. This is necessary because of
607 * the fact we are doing lazy LRU updates to minimise lock contention so the
608 * LRU does not have strict ordering. Hence we don't want to reclaim inodes
609 * with this flag set because they are the inodes that are out of order.
611 void prune_icache_sb(struct super_block
*sb
, int nr_to_scan
)
615 unsigned long reap
= 0;
617 spin_lock(&sb
->s_inode_lru_lock
);
618 for (nr_scanned
= nr_to_scan
; nr_scanned
>= 0; nr_scanned
--) {
621 if (list_empty(&sb
->s_inode_lru
))
624 inode
= list_entry(sb
->s_inode_lru
.prev
, struct inode
, i_lru
);
627 * we are inverting the sb->s_inode_lru_lock/inode->i_lock here,
628 * so use a trylock. If we fail to get the lock, just move the
629 * inode to the back of the list so we don't spin on it.
631 if (!spin_trylock(&inode
->i_lock
)) {
632 list_move(&inode
->i_lru
, &sb
->s_inode_lru
);
637 * Referenced or dirty inodes are still in use. Give them
638 * another pass through the LRU as we canot reclaim them now.
640 if (atomic_read(&inode
->i_count
) ||
641 (inode
->i_state
& ~I_REFERENCED
)) {
642 list_del_init(&inode
->i_lru
);
643 spin_unlock(&inode
->i_lock
);
644 sb
->s_nr_inodes_unused
--;
645 this_cpu_dec(nr_unused
);
649 /* recently referenced inodes get one more pass */
650 if (inode
->i_state
& I_REFERENCED
) {
651 inode
->i_state
&= ~I_REFERENCED
;
652 list_move(&inode
->i_lru
, &sb
->s_inode_lru
);
653 spin_unlock(&inode
->i_lock
);
656 if (inode_has_buffers(inode
) || inode
->i_data
.nrpages
) {
658 spin_unlock(&inode
->i_lock
);
659 spin_unlock(&sb
->s_inode_lru_lock
);
660 if (remove_inode_buffers(inode
))
661 reap
+= invalidate_mapping_pages(&inode
->i_data
,
664 spin_lock(&sb
->s_inode_lru_lock
);
666 if (inode
!= list_entry(sb
->s_inode_lru
.next
,
667 struct inode
, i_lru
))
668 continue; /* wrong inode or list_empty */
669 /* avoid lock inversions with trylock */
670 if (!spin_trylock(&inode
->i_lock
))
672 if (!can_unuse(inode
)) {
673 spin_unlock(&inode
->i_lock
);
677 WARN_ON(inode
->i_state
& I_NEW
);
678 inode
->i_state
|= I_FREEING
;
679 spin_unlock(&inode
->i_lock
);
681 list_move(&inode
->i_lru
, &freeable
);
682 sb
->s_nr_inodes_unused
--;
683 this_cpu_dec(nr_unused
);
685 if (current_is_kswapd())
686 __count_vm_events(KSWAPD_INODESTEAL
, reap
);
688 __count_vm_events(PGINODESTEAL
, reap
);
689 spin_unlock(&sb
->s_inode_lru_lock
);
691 dispose_list(&freeable
);
694 static void __wait_on_freeing_inode(struct inode
*inode
);
696 * Called with the inode lock held.
698 static struct inode
*find_inode(struct super_block
*sb
,
699 struct hlist_head
*head
,
700 int (*test
)(struct inode
*, void *),
703 struct hlist_node
*node
;
704 struct inode
*inode
= NULL
;
707 hlist_for_each_entry(inode
, node
, head
, i_hash
) {
708 spin_lock(&inode
->i_lock
);
709 if (inode
->i_sb
!= sb
) {
710 spin_unlock(&inode
->i_lock
);
713 if (!test(inode
, data
)) {
714 spin_unlock(&inode
->i_lock
);
717 if (inode
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
718 __wait_on_freeing_inode(inode
);
722 spin_unlock(&inode
->i_lock
);
729 * find_inode_fast is the fast path version of find_inode, see the comment at
730 * iget_locked for details.
732 static struct inode
*find_inode_fast(struct super_block
*sb
,
733 struct hlist_head
*head
, unsigned long ino
)
735 struct hlist_node
*node
;
736 struct inode
*inode
= NULL
;
739 hlist_for_each_entry(inode
, node
, head
, i_hash
) {
740 spin_lock(&inode
->i_lock
);
741 if (inode
->i_ino
!= ino
) {
742 spin_unlock(&inode
->i_lock
);
745 if (inode
->i_sb
!= sb
) {
746 spin_unlock(&inode
->i_lock
);
749 if (inode
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
750 __wait_on_freeing_inode(inode
);
754 spin_unlock(&inode
->i_lock
);
761 * Each cpu owns a range of LAST_INO_BATCH numbers.
762 * 'shared_last_ino' is dirtied only once out of LAST_INO_BATCH allocations,
763 * to renew the exhausted range.
765 * This does not significantly increase overflow rate because every CPU can
766 * consume at most LAST_INO_BATCH-1 unused inode numbers. So there is
767 * NR_CPUS*(LAST_INO_BATCH-1) wastage. At 4096 and 1024, this is ~0.1% of the
768 * 2^32 range, and is a worst-case. Even a 50% wastage would only increase
769 * overflow rate by 2x, which does not seem too significant.
771 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
772 * error if st_ino won't fit in target struct field. Use 32bit counter
773 * here to attempt to avoid that.
775 #define LAST_INO_BATCH 1024
776 static DEFINE_PER_CPU(unsigned int, last_ino
);
778 unsigned int get_next_ino(void)
780 unsigned int *p
= &get_cpu_var(last_ino
);
781 unsigned int res
= *p
;
784 if (unlikely((res
& (LAST_INO_BATCH
-1)) == 0)) {
785 static atomic_t shared_last_ino
;
786 int next
= atomic_add_return(LAST_INO_BATCH
, &shared_last_ino
);
788 res
= next
- LAST_INO_BATCH
;
793 put_cpu_var(last_ino
);
796 EXPORT_SYMBOL(get_next_ino
);
799 * new_inode - obtain an inode
802 * Allocates a new inode for given superblock. The default gfp_mask
803 * for allocations related to inode->i_mapping is GFP_HIGHUSER_MOVABLE.
804 * If HIGHMEM pages are unsuitable or it is known that pages allocated
805 * for the page cache are not reclaimable or migratable,
806 * mapping_set_gfp_mask() must be called with suitable flags on the
807 * newly created inode's mapping
810 struct inode
*new_inode(struct super_block
*sb
)
814 spin_lock_prefetch(&inode_sb_list_lock
);
816 inode
= alloc_inode(sb
);
818 spin_lock(&inode
->i_lock
);
820 spin_unlock(&inode
->i_lock
);
821 inode_sb_list_add(inode
);
825 EXPORT_SYMBOL(new_inode
);
828 * unlock_new_inode - clear the I_NEW state and wake up any waiters
829 * @inode: new inode to unlock
831 * Called when the inode is fully initialised to clear the new state of the
832 * inode and wake up anyone waiting for the inode to finish initialisation.
834 void unlock_new_inode(struct inode
*inode
)
836 #ifdef CONFIG_DEBUG_LOCK_ALLOC
837 if (S_ISDIR(inode
->i_mode
)) {
838 struct file_system_type
*type
= inode
->i_sb
->s_type
;
840 /* Set new key only if filesystem hasn't already changed it */
841 if (!lockdep_match_class(&inode
->i_mutex
,
842 &type
->i_mutex_key
)) {
844 * ensure nobody is actually holding i_mutex
846 mutex_destroy(&inode
->i_mutex
);
847 mutex_init(&inode
->i_mutex
);
848 lockdep_set_class(&inode
->i_mutex
,
849 &type
->i_mutex_dir_key
);
853 spin_lock(&inode
->i_lock
);
854 WARN_ON(!(inode
->i_state
& I_NEW
));
855 inode
->i_state
&= ~I_NEW
;
856 wake_up_bit(&inode
->i_state
, __I_NEW
);
857 spin_unlock(&inode
->i_lock
);
859 EXPORT_SYMBOL(unlock_new_inode
);
862 * iget5_locked - obtain an inode from a mounted file system
863 * @sb: super block of file system
864 * @hashval: hash value (usually inode number) to get
865 * @test: callback used for comparisons between inodes
866 * @set: callback used to initialize a new struct inode
867 * @data: opaque data pointer to pass to @test and @set
869 * Search for the inode specified by @hashval and @data in the inode cache,
870 * and if present it is return it with an increased reference count. This is
871 * a generalized version of iget_locked() for file systems where the inode
872 * number is not sufficient for unique identification of an inode.
874 * If the inode is not in cache, allocate a new inode and return it locked,
875 * hashed, and with the I_NEW flag set. The file system gets to fill it in
876 * before unlocking it via unlock_new_inode().
878 * Note both @test and @set are called with the inode_hash_lock held, so can't
881 struct inode
*iget5_locked(struct super_block
*sb
, unsigned long hashval
,
882 int (*test
)(struct inode
*, void *),
883 int (*set
)(struct inode
*, void *), void *data
)
885 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
888 spin_lock(&inode_hash_lock
);
889 inode
= find_inode(sb
, head
, test
, data
);
890 spin_unlock(&inode_hash_lock
);
893 wait_on_inode(inode
);
897 inode
= alloc_inode(sb
);
901 spin_lock(&inode_hash_lock
);
902 /* We released the lock, so.. */
903 old
= find_inode(sb
, head
, test
, data
);
905 if (set(inode
, data
))
908 spin_lock(&inode
->i_lock
);
909 inode
->i_state
= I_NEW
;
910 hlist_add_head(&inode
->i_hash
, head
);
911 spin_unlock(&inode
->i_lock
);
912 inode_sb_list_add(inode
);
913 spin_unlock(&inode_hash_lock
);
915 /* Return the locked inode with I_NEW set, the
916 * caller is responsible for filling in the contents
922 * Uhhuh, somebody else created the same inode under
923 * us. Use the old inode instead of the one we just
926 spin_unlock(&inode_hash_lock
);
927 destroy_inode(inode
);
929 wait_on_inode(inode
);
934 spin_unlock(&inode_hash_lock
);
935 destroy_inode(inode
);
938 EXPORT_SYMBOL(iget5_locked
);
941 * iget_locked - obtain an inode from a mounted file system
942 * @sb: super block of file system
943 * @ino: inode number to get
945 * Search for the inode specified by @ino in the inode cache and if present
946 * return it with an increased reference count. This is for file systems
947 * where the inode number is sufficient for unique identification of an inode.
949 * If the inode is not in cache, allocate a new inode and return it locked,
950 * hashed, and with the I_NEW flag set. The file system gets to fill it in
951 * before unlocking it via unlock_new_inode().
953 struct inode
*iget_locked(struct super_block
*sb
, unsigned long ino
)
955 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
958 spin_lock(&inode_hash_lock
);
959 inode
= find_inode_fast(sb
, head
, ino
);
960 spin_unlock(&inode_hash_lock
);
962 wait_on_inode(inode
);
966 inode
= alloc_inode(sb
);
970 spin_lock(&inode_hash_lock
);
971 /* We released the lock, so.. */
972 old
= find_inode_fast(sb
, head
, ino
);
975 spin_lock(&inode
->i_lock
);
976 inode
->i_state
= I_NEW
;
977 hlist_add_head(&inode
->i_hash
, head
);
978 spin_unlock(&inode
->i_lock
);
979 inode_sb_list_add(inode
);
980 spin_unlock(&inode_hash_lock
);
982 /* Return the locked inode with I_NEW set, the
983 * caller is responsible for filling in the contents
989 * Uhhuh, somebody else created the same inode under
990 * us. Use the old inode instead of the one we just
993 spin_unlock(&inode_hash_lock
);
994 destroy_inode(inode
);
996 wait_on_inode(inode
);
1000 EXPORT_SYMBOL(iget_locked
);
1003 * search the inode cache for a matching inode number.
1004 * If we find one, then the inode number we are trying to
1005 * allocate is not unique and so we should not use it.
1007 * Returns 1 if the inode number is unique, 0 if it is not.
1009 static int test_inode_iunique(struct super_block
*sb
, unsigned long ino
)
1011 struct hlist_head
*b
= inode_hashtable
+ hash(sb
, ino
);
1012 struct hlist_node
*node
;
1013 struct inode
*inode
;
1015 spin_lock(&inode_hash_lock
);
1016 hlist_for_each_entry(inode
, node
, b
, i_hash
) {
1017 if (inode
->i_ino
== ino
&& inode
->i_sb
== sb
) {
1018 spin_unlock(&inode_hash_lock
);
1022 spin_unlock(&inode_hash_lock
);
1028 * iunique - get a unique inode number
1030 * @max_reserved: highest reserved inode number
1032 * Obtain an inode number that is unique on the system for a given
1033 * superblock. This is used by file systems that have no natural
1034 * permanent inode numbering system. An inode number is returned that
1035 * is higher than the reserved limit but unique.
1038 * With a large number of inodes live on the file system this function
1039 * currently becomes quite slow.
1041 ino_t
iunique(struct super_block
*sb
, ino_t max_reserved
)
1044 * On a 32bit, non LFS stat() call, glibc will generate an EOVERFLOW
1045 * error if st_ino won't fit in target struct field. Use 32bit counter
1046 * here to attempt to avoid that.
1048 static DEFINE_SPINLOCK(iunique_lock
);
1049 static unsigned int counter
;
1052 spin_lock(&iunique_lock
);
1054 if (counter
<= max_reserved
)
1055 counter
= max_reserved
+ 1;
1057 } while (!test_inode_iunique(sb
, res
));
1058 spin_unlock(&iunique_lock
);
1062 EXPORT_SYMBOL(iunique
);
1064 struct inode
*igrab(struct inode
*inode
)
1066 spin_lock(&inode
->i_lock
);
1067 if (!(inode
->i_state
& (I_FREEING
|I_WILL_FREE
))) {
1069 spin_unlock(&inode
->i_lock
);
1071 spin_unlock(&inode
->i_lock
);
1073 * Handle the case where s_op->clear_inode is not been
1074 * called yet, and somebody is calling igrab
1075 * while the inode is getting freed.
1081 EXPORT_SYMBOL(igrab
);
1084 * ilookup5_nowait - search for an inode in the inode cache
1085 * @sb: super block of file system to search
1086 * @hashval: hash value (usually inode number) to search for
1087 * @test: callback used for comparisons between inodes
1088 * @data: opaque data pointer to pass to @test
1090 * Search for the inode specified by @hashval and @data in the inode cache.
1091 * If the inode is in the cache, the inode is returned with an incremented
1094 * Note: I_NEW is not waited upon so you have to be very careful what you do
1095 * with the returned inode. You probably should be using ilookup5() instead.
1097 * Note2: @test is called with the inode_hash_lock held, so can't sleep.
1099 struct inode
*ilookup5_nowait(struct super_block
*sb
, unsigned long hashval
,
1100 int (*test
)(struct inode
*, void *), void *data
)
1102 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1103 struct inode
*inode
;
1105 spin_lock(&inode_hash_lock
);
1106 inode
= find_inode(sb
, head
, test
, data
);
1107 spin_unlock(&inode_hash_lock
);
1111 EXPORT_SYMBOL(ilookup5_nowait
);
1114 * ilookup5 - search for an inode in the inode cache
1115 * @sb: super block of file system to search
1116 * @hashval: hash value (usually inode number) to search for
1117 * @test: callback used for comparisons between inodes
1118 * @data: opaque data pointer to pass to @test
1120 * Search for the inode specified by @hashval and @data in the inode cache,
1121 * and if the inode is in the cache, return the inode with an incremented
1122 * reference count. Waits on I_NEW before returning the inode.
1123 * returned with an incremented reference count.
1125 * This is a generalized version of ilookup() for file systems where the
1126 * inode number is not sufficient for unique identification of an inode.
1128 * Note: @test is called with the inode_hash_lock held, so can't sleep.
1130 struct inode
*ilookup5(struct super_block
*sb
, unsigned long hashval
,
1131 int (*test
)(struct inode
*, void *), void *data
)
1133 struct inode
*inode
= ilookup5_nowait(sb
, hashval
, test
, data
);
1136 wait_on_inode(inode
);
1139 EXPORT_SYMBOL(ilookup5
);
1142 * ilookup - search for an inode in the inode cache
1143 * @sb: super block of file system to search
1144 * @ino: inode number to search for
1146 * Search for the inode @ino in the inode cache, and if the inode is in the
1147 * cache, the inode is returned with an incremented reference count.
1149 struct inode
*ilookup(struct super_block
*sb
, unsigned long ino
)
1151 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1152 struct inode
*inode
;
1154 spin_lock(&inode_hash_lock
);
1155 inode
= find_inode_fast(sb
, head
, ino
);
1156 spin_unlock(&inode_hash_lock
);
1159 wait_on_inode(inode
);
1162 EXPORT_SYMBOL(ilookup
);
1164 int insert_inode_locked(struct inode
*inode
)
1166 struct super_block
*sb
= inode
->i_sb
;
1167 ino_t ino
= inode
->i_ino
;
1168 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, ino
);
1171 struct hlist_node
*node
;
1172 struct inode
*old
= NULL
;
1173 spin_lock(&inode_hash_lock
);
1174 hlist_for_each_entry(old
, node
, head
, i_hash
) {
1175 if (old
->i_ino
!= ino
)
1177 if (old
->i_sb
!= sb
)
1179 spin_lock(&old
->i_lock
);
1180 if (old
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
1181 spin_unlock(&old
->i_lock
);
1186 if (likely(!node
)) {
1187 spin_lock(&inode
->i_lock
);
1188 inode
->i_state
|= I_NEW
;
1189 hlist_add_head(&inode
->i_hash
, head
);
1190 spin_unlock(&inode
->i_lock
);
1191 spin_unlock(&inode_hash_lock
);
1195 spin_unlock(&old
->i_lock
);
1196 spin_unlock(&inode_hash_lock
);
1198 if (unlikely(!inode_unhashed(old
))) {
1205 EXPORT_SYMBOL(insert_inode_locked
);
1207 int insert_inode_locked4(struct inode
*inode
, unsigned long hashval
,
1208 int (*test
)(struct inode
*, void *), void *data
)
1210 struct super_block
*sb
= inode
->i_sb
;
1211 struct hlist_head
*head
= inode_hashtable
+ hash(sb
, hashval
);
1214 struct hlist_node
*node
;
1215 struct inode
*old
= NULL
;
1217 spin_lock(&inode_hash_lock
);
1218 hlist_for_each_entry(old
, node
, head
, i_hash
) {
1219 if (old
->i_sb
!= sb
)
1221 if (!test(old
, data
))
1223 spin_lock(&old
->i_lock
);
1224 if (old
->i_state
& (I_FREEING
|I_WILL_FREE
)) {
1225 spin_unlock(&old
->i_lock
);
1230 if (likely(!node
)) {
1231 spin_lock(&inode
->i_lock
);
1232 inode
->i_state
|= I_NEW
;
1233 hlist_add_head(&inode
->i_hash
, head
);
1234 spin_unlock(&inode
->i_lock
);
1235 spin_unlock(&inode_hash_lock
);
1239 spin_unlock(&old
->i_lock
);
1240 spin_unlock(&inode_hash_lock
);
1242 if (unlikely(!inode_unhashed(old
))) {
1249 EXPORT_SYMBOL(insert_inode_locked4
);
1252 int generic_delete_inode(struct inode
*inode
)
1256 EXPORT_SYMBOL(generic_delete_inode
);
1259 * Normal UNIX filesystem behaviour: delete the
1260 * inode when the usage count drops to zero, and
1263 int generic_drop_inode(struct inode
*inode
)
1265 return !inode
->i_nlink
|| inode_unhashed(inode
);
1267 EXPORT_SYMBOL_GPL(generic_drop_inode
);
1270 * Called when we're dropping the last reference
1273 * Call the FS "drop_inode()" function, defaulting to
1274 * the legacy UNIX filesystem behaviour. If it tells
1275 * us to evict inode, do so. Otherwise, retain inode
1276 * in cache if fs is alive, sync and evict if fs is
1279 static void iput_final(struct inode
*inode
)
1281 struct super_block
*sb
= inode
->i_sb
;
1282 const struct super_operations
*op
= inode
->i_sb
->s_op
;
1285 WARN_ON(inode
->i_state
& I_NEW
);
1288 drop
= op
->drop_inode(inode
);
1290 drop
= generic_drop_inode(inode
);
1292 if (!drop
&& (sb
->s_flags
& MS_ACTIVE
)) {
1293 inode
->i_state
|= I_REFERENCED
;
1294 if (!(inode
->i_state
& (I_DIRTY
|I_SYNC
)))
1295 inode_lru_list_add(inode
);
1296 spin_unlock(&inode
->i_lock
);
1301 inode
->i_state
|= I_WILL_FREE
;
1302 spin_unlock(&inode
->i_lock
);
1303 write_inode_now(inode
, 1);
1304 spin_lock(&inode
->i_lock
);
1305 WARN_ON(inode
->i_state
& I_NEW
);
1306 inode
->i_state
&= ~I_WILL_FREE
;
1309 inode
->i_state
|= I_FREEING
;
1310 inode_lru_list_del(inode
);
1311 spin_unlock(&inode
->i_lock
);
1317 * iput - put an inode
1318 * @inode: inode to put
1320 * Puts an inode, dropping its usage count. If the inode use count hits
1321 * zero, the inode is then freed and may also be destroyed.
1323 * Consequently, iput() can sleep.
1325 void iput(struct inode
*inode
)
1328 BUG_ON(inode
->i_state
& I_CLEAR
);
1330 if (atomic_dec_and_lock(&inode
->i_count
, &inode
->i_lock
))
1334 EXPORT_SYMBOL(iput
);
1337 * bmap - find a block number in a file
1338 * @inode: inode of file
1339 * @block: block to find
1341 * Returns the block number on the device holding the inode that
1342 * is the disk block number for the block of the file requested.
1343 * That is, asked for block 4 of inode 1 the function will return the
1344 * disk block relative to the disk start that holds that block of the
1347 sector_t
bmap(struct inode
*inode
, sector_t block
)
1350 if (inode
->i_mapping
->a_ops
->bmap
)
1351 res
= inode
->i_mapping
->a_ops
->bmap(inode
->i_mapping
, block
);
1354 EXPORT_SYMBOL(bmap
);
1357 * With relative atime, only update atime if the previous atime is
1358 * earlier than either the ctime or mtime or if at least a day has
1359 * passed since the last atime update.
1361 static int relatime_need_update(struct vfsmount
*mnt
, struct inode
*inode
,
1362 struct timespec now
)
1365 if (!(mnt
->mnt_flags
& MNT_RELATIME
))
1368 * Is mtime younger than atime? If yes, update atime:
1370 if (timespec_compare(&inode
->i_mtime
, &inode
->i_atime
) >= 0)
1373 * Is ctime younger than atime? If yes, update atime:
1375 if (timespec_compare(&inode
->i_ctime
, &inode
->i_atime
) >= 0)
1379 * Is the previous atime value older than a day? If yes,
1382 if ((long)(now
.tv_sec
- inode
->i_atime
.tv_sec
) >= 24*60*60)
1385 * Good, we can skip the atime update:
1391 * touch_atime - update the access time
1392 * @mnt: mount the inode is accessed on
1393 * @dentry: dentry accessed
1395 * Update the accessed time on an inode and mark it for writeback.
1396 * This function automatically handles read only file systems and media,
1397 * as well as the "noatime" flag and inode specific "noatime" markers.
1399 void touch_atime(struct vfsmount
*mnt
, struct dentry
*dentry
)
1401 struct inode
*inode
= dentry
->d_inode
;
1402 struct timespec now
;
1404 if (inode
->i_flags
& S_NOATIME
)
1406 if (IS_NOATIME(inode
))
1408 if ((inode
->i_sb
->s_flags
& MS_NODIRATIME
) && S_ISDIR(inode
->i_mode
))
1411 if (mnt
->mnt_flags
& MNT_NOATIME
)
1413 if ((mnt
->mnt_flags
& MNT_NODIRATIME
) && S_ISDIR(inode
->i_mode
))
1416 now
= current_fs_time(inode
->i_sb
);
1418 if (!relatime_need_update(mnt
, inode
, now
))
1421 if (timespec_equal(&inode
->i_atime
, &now
))
1424 if (mnt_want_write(mnt
))
1427 inode
->i_atime
= now
;
1428 mark_inode_dirty_sync(inode
);
1429 mnt_drop_write(mnt
);
1431 EXPORT_SYMBOL(touch_atime
);
1434 * file_update_time - update mtime and ctime time
1435 * @file: file accessed
1437 * Update the mtime and ctime members of an inode and mark the inode
1438 * for writeback. Note that this function is meant exclusively for
1439 * usage in the file write path of filesystems, and filesystems may
1440 * choose to explicitly ignore update via this function with the
1441 * S_NOCMTIME inode flag, e.g. for network filesystem where these
1442 * timestamps are handled by the server.
1445 void file_update_time(struct file
*file
)
1447 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1448 struct timespec now
;
1449 enum { S_MTIME
= 1, S_CTIME
= 2, S_VERSION
= 4 } sync_it
= 0;
1451 /* First try to exhaust all avenues to not sync */
1452 if (IS_NOCMTIME(inode
))
1455 now
= current_fs_time(inode
->i_sb
);
1456 if (!timespec_equal(&inode
->i_mtime
, &now
))
1459 if (!timespec_equal(&inode
->i_ctime
, &now
))
1462 if (IS_I_VERSION(inode
))
1463 sync_it
|= S_VERSION
;
1468 /* Finally allowed to write? Takes lock. */
1469 if (mnt_want_write_file(file
))
1472 /* Only change inode inside the lock region */
1473 if (sync_it
& S_VERSION
)
1474 inode_inc_iversion(inode
);
1475 if (sync_it
& S_CTIME
)
1476 inode
->i_ctime
= now
;
1477 if (sync_it
& S_MTIME
)
1478 inode
->i_mtime
= now
;
1479 mark_inode_dirty_sync(inode
);
1480 mnt_drop_write(file
->f_path
.mnt
);
1482 EXPORT_SYMBOL(file_update_time
);
1484 int inode_needs_sync(struct inode
*inode
)
1488 if (S_ISDIR(inode
->i_mode
) && IS_DIRSYNC(inode
))
1492 EXPORT_SYMBOL(inode_needs_sync
);
1494 int inode_wait(void *word
)
1499 EXPORT_SYMBOL(inode_wait
);
1502 * If we try to find an inode in the inode hash while it is being
1503 * deleted, we have to wait until the filesystem completes its
1504 * deletion before reporting that it isn't found. This function waits
1505 * until the deletion _might_ have completed. Callers are responsible
1506 * to recheck inode state.
1508 * It doesn't matter if I_NEW is not set initially, a call to
1509 * wake_up_bit(&inode->i_state, __I_NEW) after removing from the hash list
1512 static void __wait_on_freeing_inode(struct inode
*inode
)
1514 wait_queue_head_t
*wq
;
1515 DEFINE_WAIT_BIT(wait
, &inode
->i_state
, __I_NEW
);
1516 wq
= bit_waitqueue(&inode
->i_state
, __I_NEW
);
1517 prepare_to_wait(wq
, &wait
.wait
, TASK_UNINTERRUPTIBLE
);
1518 spin_unlock(&inode
->i_lock
);
1519 spin_unlock(&inode_hash_lock
);
1521 finish_wait(wq
, &wait
.wait
);
1522 spin_lock(&inode_hash_lock
);
1525 static __initdata
unsigned long ihash_entries
;
1526 static int __init
set_ihash_entries(char *str
)
1530 ihash_entries
= simple_strtoul(str
, &str
, 0);
1533 __setup("ihash_entries=", set_ihash_entries
);
1536 * Initialize the waitqueues and inode hash table.
1538 void __init
inode_init_early(void)
1542 /* If hashes are distributed across NUMA nodes, defer
1543 * hash allocation until vmalloc space is available.
1549 alloc_large_system_hash("Inode-cache",
1550 sizeof(struct hlist_head
),
1558 for (loop
= 0; loop
< (1 << i_hash_shift
); loop
++)
1559 INIT_HLIST_HEAD(&inode_hashtable
[loop
]);
1562 void __init
inode_init(void)
1566 /* inode slab cache */
1567 inode_cachep
= kmem_cache_create("inode_cache",
1568 sizeof(struct inode
),
1570 (SLAB_RECLAIM_ACCOUNT
|SLAB_PANIC
|
1574 /* Hash may have been set up in inode_init_early */
1579 alloc_large_system_hash("Inode-cache",
1580 sizeof(struct hlist_head
),
1588 for (loop
= 0; loop
< (1 << i_hash_shift
); loop
++)
1589 INIT_HLIST_HEAD(&inode_hashtable
[loop
]);
1592 void init_special_inode(struct inode
*inode
, umode_t mode
, dev_t rdev
)
1594 inode
->i_mode
= mode
;
1595 if (S_ISCHR(mode
)) {
1596 inode
->i_fop
= &def_chr_fops
;
1597 inode
->i_rdev
= rdev
;
1598 } else if (S_ISBLK(mode
)) {
1599 inode
->i_fop
= &def_blk_fops
;
1600 inode
->i_rdev
= rdev
;
1601 } else if (S_ISFIFO(mode
))
1602 inode
->i_fop
= &def_fifo_fops
;
1603 else if (S_ISSOCK(mode
))
1604 inode
->i_fop
= &bad_sock_fops
;
1606 printk(KERN_DEBUG
"init_special_inode: bogus i_mode (%o) for"
1607 " inode %s:%lu\n", mode
, inode
->i_sb
->s_id
,
1610 EXPORT_SYMBOL(init_special_inode
);
1613 * inode_init_owner - Init uid,gid,mode for new inode according to posix standards
1615 * @dir: Directory inode
1616 * @mode: mode of the new inode
1618 void inode_init_owner(struct inode
*inode
, const struct inode
*dir
,
1621 inode
->i_uid
= current_fsuid();
1622 if (dir
&& dir
->i_mode
& S_ISGID
) {
1623 inode
->i_gid
= dir
->i_gid
;
1627 inode
->i_gid
= current_fsgid();
1628 inode
->i_mode
= mode
;
1630 EXPORT_SYMBOL(inode_init_owner
);
1633 * inode_owner_or_capable - check current task permissions to inode
1634 * @inode: inode being checked
1636 * Return true if current either has CAP_FOWNER to the inode, or
1639 bool inode_owner_or_capable(const struct inode
*inode
)
1641 struct user_namespace
*ns
= inode_userns(inode
);
1643 if (current_user_ns() == ns
&& current_fsuid() == inode
->i_uid
)
1645 if (ns_capable(ns
, CAP_FOWNER
))
1649 EXPORT_SYMBOL(inode_owner_or_capable
);